bio-rs
bio-rs is a Rust-based bio-AI data/tooling engine for validating biological data, preparing model-ready inputs, and emitting reproducible JSON contracts across CLI, Rust, Python, WASM, local services, and agent tools.
What It Does
- Validates protein, DNA, RNA, FASTQ, PDB, SMILES, SDF, and MOL2 inputs.
- Builds tokenizer IDs, attention masks, model-input payloads, reports, and pipeline locks.
- Checks package manifests, model artifacts, checksums, fixtures, and runtime bridge plans.
- Exposes the same local-first contracts through the CLI, Rust, Python, WASM, MCP, and local HTTP/service schemas.
See docs/sequence-kind-support.md before making broad DNA/RNA support claims. Package skeleton generation and Python/Hugging Face conversion remain protein-first.
Quickstart
|
More examples: docs/quickstart.md
Evidence
Committed benchmark reports track release regressions across CLI, Python, WASM, MCP, and optional backend smoke paths. They are reproducible guardrails, not universal throughput claims.
Current checked-in examples include:
- CLI DNA/RNA validation on 256 records in about 6 ms: benchmarks/cli_surfaces.md
- WASM workflow on 256 records in about 8-10 ms: benchmarks/wasm_bindings.md
- MCP doctor request overhead at about 60 us: benchmarks/mcp_server.md
Main Docs
- Installation
- CLI contract
- JSON schemas
- Protein, DNA, and RNA support
- Biological formats
- Structure support
- Molecule support
- Package format and conversion
- Pipeline config
- Service interface and local HTTP mode
- Rust API
- Python API
- WASM API
- Candle backend
- Versioning policy
- Error codes
- Citation metadata
Development
License
Dual licensed under MIT OR Apache-2.0. If you use bio-rs in research software or publications, cite the repository and version via CITATION.cff.